Embedded Computer Vision Safety System for Freight Elevators Using SSD-MobileNet

Abstract

This paper presents the design and implementation of an intelligent safety system for freight elevators based on computer vision, aimed at reducing workplace accidents caused by improper use of such equipment. The proposed system relies on an SSD-MobileNet convolutional neural network, trained with a dataset of 1,050 labeled images under varying lighting conditions and deployed on a low-cost ESP32-CAM microcontroller. The system detects the presence of individuals at the elevator entrance and, through communication with a Siemens PLC S7-1200 and a variable frequency drive (VFD), determines whether to enable or block motor activation. Validation was conducted in a controlled laboratory environment using a three-level platform with a 25 kg load. The experimental results yielded an F1-score of 93.13%, a recall of 90.63%, and a specificity of 96.00%. The system is presented as a functional proof of concept with future potential for deployment in real industrial environments, highlighting its low cost, effective integration, and preventive approach to occupational safety.